AI Skills Top Employer Priorities—Only 29% Claim Mature Capabilities

On October 8, 2026, edX Enterprise released The State of Enterprise Learning 2026 from Arlington, Virginia: 83% of surveyed workforce leaders called AI capability an active priority, but only 29% reported mature AI capabilities in their organizations, and just 26% were highly confident their learning and development function could deliver the capabilities workers need. The gap puts an immediate question behind employers’ training plans: whether a declared priority is becoming a dependable capability across the business.

The NewtonX account of the research describes an online survey of 507 verified learning and workforce strategy leaders, fielded from June to July 2026 at organizations with at least 500 employees across the Americas, Europe, the Middle East, Africa and Asia-Pacific; it also puts the share of learning functions with executive backing, clear AI skill priorities, and budget and staffing at 22%. A smaller group, 15% of the sample, met a combined threshold for mature AI capability and high confidence in learning delivery. These figures reflect leaders’ assessments of their organizations, not independently tested employee skills.

Priority needs an executive mandate

A stated priority gives a learning team a direction; a mandate gives it authority and capacity. According to 2U’s release of the findings, 78% of respondents lacked a strong AI upskilling mandate, while NewtonX analyst Chris Kelly cautioned that deciding whether HR, L&D or IT owns the work “alone builds nothing.” A named owner may still lack agreement on the capabilities the business needs, a budget to develop them or enough staff to deliver learning at scale.

Organizations in the combined maturity and delivery-confidence group were 3.9 times as likely as others to have a strong mandate. That is an association within the survey, rather than evidence that issuing a mandate will cause capability to improve. It does make the reported bottleneck more specific than a shortage of AI courses: the learning function needs a defined assignment, executive support and resources aligned with that assignment.

In an analysis of factors associated with membership in that group, mandate strength accounted for 59% of the factors’ combined influence and measurement depth for 34%; formal ownership accounted for 1%. Those figures describe the analysis, not the expected effect of changing a policy. They suggest that settling a departmental ownership debate has limited value if the chosen team still lacks a clear goal and the means to pursue it.

Measurement separates activity from capability

The stronger group tracked 1.6 times as many measures of learning success as other respondents, including outcomes beyond course participation. Enrollment and completion establish that training reached people. They offer a narrower view of whether workers can use AI in their roles, or whether learning addresses a capability the organization needs. A deeper measure connects instruction to those target capabilities and to evidence of their use at work.

This distinction changes what a successful rollout would mean. A course can be well attended even when its content is disconnected from role expectations or when managers have no basis for recognizing improved performance. Outcome measures give a learning function more information about where a program is working and where its reach ends. The survey establishes an association between broader measurement and the stronger group; it does not show that adding metrics by itself produces mature AI capability.

Leadership development remains on a separate track

The stronger organizations integrated AI learning with leadership development at 2.7 times the rate of the rest of the sample, while 85% of all respondents did not develop the two as one connected effort. That separation matters because employees can learn tools while managers continue to set expectations, assess work and guide teams without a shared understanding of how AI fits those responsibilities.

Connecting the programs would make the intended workplace change more explicit. Leaders need to know which uses of AI belong in a role and how to judge work shaped by those tools; employees need expectations that match what their training prepares them to do. Integration was more common in the stronger group, but the survey cannot establish whether it caused the difference. The finding identifies a reported practice associated with maturity, alongside mandate strength and measurement.

The narrower threshold for maturity

The combined threshold joins two assessments that can diverge: established AI capability and confidence that learning and development can keep delivering what the workforce needs. An organization may regard AI as urgent and may have training underway while remaining uncertain about its ability to build the required capabilities consistently. That delivery question helps explain why reported maturity alone does not place an organization in the smaller leading group.

The findings describe a progression from executive direction to budget and staffing, then to measures of capability, leadership expectations and confidence in delivery. This is an interpretation of the reported practices, not a tested formula or a ranking of employers. It also keeps the survey’s limit in view: senior learning and workforce leaders assessed their own organizations, and the research did not measure employees’ AI skills directly.

For employers comparing their programs with these results, a larger course catalog could raise participation without changing the underlying assessment of readiness. The consequential question is whether the organization can support the training, see capability develop in work and sustain delivery as roles change.

Also read:

• Together Link Swaps Coding Models—but Its Savings Claim Needs Testing